In the realm of actuarial science, Akur8 has made a significant stride with the release of their latest research paper titled “Derivative Lasso: Credibility-Based Signal Fitting for GLMs.”
The paper, crafted by Akur8’s actuarial experts Mattia Casotto and Thomas Holmes, tackles the longstanding challenge faced by actuaries: the trade-off between creating transparent, but complex Generalized Linear Models (GLMs) and opting for automated, yet opaque model-building techniques.
Generalized Linear Models have been the cornerstone of actuarial risk assessment for over two decades, thanks to their robust statistical framework that allows for clear-cut assumptions and direct output for rating tables. Despite the advent of advanced machine learning models like GBMs and Random Forests, their lack of transparency has remained a hurdle for actuaries who need to adjust and interact with their predictive models.
Akur8’s Mattia Casotto sheds light on the innovative approach detailed in the paper, emphasizing the derivative lasso’s ability to integrate credibility aspects into GLMs while preserving their standard framework. This technique allows actuaries to effectively model nonlinearities without significant alterations to the GLM optimization process.
Samuel Falmagne, the Co-founder and CEO of Akur8, expresses pride in his team’s collaborative effort to enrich the actuarial literature with this comprehensive research. Akur8’s solution stands out for its user-friendly approach to automating risk and rate modeling, providing insurers with a host of benefits. These include time savings on data preparation and modeling, transparent GLM outputs, and expedited market readiness—all achieved without the need for coding expertise.
Read the whitepaper here.
Keep up with all the latest FinTech news here
Copyright © 2023 FinTech Global
